Documentation: https://github.com/openai/whisper?utm_source=chatgpt.com

Overview

Whisper, simple to use, a little bit more demanding to set up. Let's take a look at the dependencies below:

Whisper
   │
   ├── PyTorch
   │      └── Visual C++ Runtime
   │
   └── FFmpeg

Installation

Visual C++ Runtime

Get-WmiObject -Class Win32_Product | Where-Object {$_.Name -like "*Visual C++*"} | Select Name

Confirm that you have the following installed:

• Microsoft Visual C++ 2022 X64 Minimum Runtime

• Microsoft Visual C++ 2022 X64 Additional Runtime

If not:

• In https://learn.microsoft.com/en-us/cpp/windows/latest-supported-vc-redist?view=msvc-170
  Download and Install: Latest supported Redistributable version

PyTorch

(To avoid conflicts with other Python packages, it is recommended to create a dedicated virtual environment for this project)
 

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

Confirm the installation:

python -c "import torch; print(torch.__version__)"

The expected output will be something like:
2.x.x+cpu

FFmpeg

Whisper uses FFmpeg to decode and read audio files. Check whether FFmpeg is already installed:

ffmpeg -version

If the command is not recognized, you will need to install it:

From https://www.gyan.dev/ffmpeg/builds/, download: ffmpeg-git-essentials.7z

Create a folder named ffmpeg in C:/ and drag the content of the download to ffmpeg:

C:/
└── ffmpeg/
    ├── bin/
    │   ├── ffmpeg.exe
    │   ├── ffplay.exe
    │   └── ffprobe.exe
    ├── doc/
    └── LICENSE

Adding ffmpeg to PATH:

Environment Variables > Edit Environment Variables > User Variables > PATH > New > C:/ffmpeg/bin
 

Whisper

Installation

pip install openai-whisper

Usage

whisper audio.mp3 --model base --language English

Available models

Whisper provides multiple models, allowing you to balance speed and accuracy:

• tiny
• base
• small
• medium
• large
• turbo
  ​​​

Larger models generally offer better accuracy at the cost of higher processing requirements. 
I've tried it on multiple audios with the base model and did a good job with the English language.

Language support

The same flexibility applies to languages. Whisper supports a wide range of languages, which can be explicitly specified or automatically detected, making it suitable for multilingual transcription workflows.

When used correctly, this tool can be a real game changer.
Long hours of meetings, interviews, or conversations can be automatically transcribed by a computer and later summarised, all with the help of AI.